产业组织理论前沿

RecentDevelopmentsinIndustrialEconomics:

ASelectiveReview

GuofuTan

Firstversion,September,2004;thisversion,March2005

Abstract.Thispaperprovidesaselectivereviewoftherecentindus-

trialeconomicsliterature,withafocusontheeconomicsofnetworksandthe

theoryofmarketdesign.Italsodiscussessomeoftheimportantmethodolog-

icalcontributionsofgametheorytotheanalysisofbusinessstrategies,public

policiesandtheperformanceofmarketmechanisms.

Contents

1Introduction...................................2GameTheoryandIndustrialEconomics...................3NetworkIndustries...............................4MarketDesign..................................5ForeignDirectInvestment...........................

6References....................................2413253236GuofuTanisaProfessorofEconomicsattheUniversityofSouthernCalifornia,LosAngeles,andaSeniorConsultantatDeltaEconomicsGroup,Inc.,Vancouver.TheauthorthanksZhiqiChen,MarcDuhamel,HengJu,GuillaumeRoger,ananonymousreferee,andparticipantsattheworkshopon“IndustrialEconomicsandPerformanceinCanada”forusefulcommentsandsugges-tionsandtheIndustryCanadafor nancialsupport.

1

1.Introduction

GoogleInc.,currentlyoneofthemostpopularcompanies,recentlywentpublic.Itsinitialpublico ering(IPO)processhasgeneratedmuchdebateamonganalysts,investorsandthelike.Twoofthemanyissuesdiscussedaremostinterestingfromtheperspectiveofindustrialeconomics.OneconcernsGoogle’snewIPOmechanismandtheotherisitsbusinessmodel.

InsteadoffollowingthecustomaryIPOprocess,Googleusedanauctionmecha-nism,calledDutchauction(oruniform-priceauction),toallocateitsequityshares.TheusualIPOmechanisminvolvesagroupofinvestmentbankshelpingthecompanytosetaprice,typicallylow,andinitiallyallocatingsharestotheirpreferredclients(mostlyinstitutionalclientssuchaspensionandmutualfunds,andinsurancecom-panies)inexchangeforhighfees.Smallinvestorsrarelygetachancetobuysharesinthisprocess.Undertheauctionmechanism,investors(smallorlarge)submitbidsspecifyingapriceandaquantitydesired.Thepriceisultimatelydeterminedbydemand(thesumofallbids)andsupply.Thismarketclearingpriceshouldinprin-ciplebehigherthantheonesetbyinvestmentbanks,raisingmoremoneythanthetraditionalmechanisms.Anauctionalsoallowsthecompanytoopenits otationtoabroadbaseofindividualshareholdersandpotentiallypreventsasuddenpriceriseonthe rstdayoftradingthatoftenoccursfollowingthetraditionalIPOs.Ontheotherhand,theremaybecaveatstotheauctionmechanismtooifitisnotimple-mentedcarefully.Thecompanyandinvestorsneedtobewaretheriskofbiddingaswellasthepresenceofthe“winner’scurse”associatedwiththeauctionmechanism.1IstheDutchauctionabettermechanismforIPOsthanthetraditionalones?Howtoimproveauctiondesignandreducetheimpactofthewinner’scurse?

OnthebusinesssideGoogle’smodelpresentsmanyinterestingfeatures,oneofwhichrelatestoGoogle’stwodistinctgroupsofcustomers:itsusersandadvertise-mentsponsors.WhileitsWebsiteandsearchenginearefreeforallusers,GooglegetsmostofitsrevenuefromInternetadvertisingprimarilythroughsponsoredlinksthatcomeupwithanysearchbyusers.Asaplatformprovider,Googlebringsthese

Intheirarticle“GoogleShouldBewarethe‘Winner’sCurse’,”FinancialTimes,May3,2004,three-and-a-halfmonthsbeforeGooglewentpublic,EricBudishandPaulKlempererpointedoutthedangerofitsDutchauctiondesignandtheriskoftheWinner’sCurse.Anothereconomist,BarryNalebu ,warnedthatitmightnotbeintheinvestors’bestinteresttoparticipateinGoogle’sauctiongame,notjustbecauseoftheWinner’sCursebutalsobecauseGooglehastheadvantageofgettingtoseeallthebidsbeforedecidingwhentoendtheauction;see“GoingOnce?GoingTwice?LetitGo...”,WashingtonPost,August8,2004.Inanauctionenvironmentwherebiddershavetheirownestimatesofacommonbutunknownvalueoftheitemforsale,thebidderwhomostoverestimatesthevaluewilltendtowinintheauction.Arationalbiddershouldtakethisintoaccountandavoidthewinner’scursebycurtailingtheirbid.1

twogroupsofparticipantsonboardandchargesonegroupwhilesubsidizingtheotherone.Thistypeofcross-subsidizationstrategiesalsoarisesinotherbusinessessuchasnewspapers,televisionnetworks,yellowpages,videogames,shoppingmall,paymentcardsystems,andsoon.Thesemarkets,calledtwo-sided(ormoregener-allymulti-sided)marketsbyindustrialeconomists,shareacommonfeaturethattwoormoregroupsofindividualsneedtointeractthroughintermediariesorplatformproviders.InthepresenceofincreasingcompetitionfromproviderssuchasYahoo!andMicrosoft,choosingtherightpricingstrategymaybecomemorechallengingforGoogle.

Anotherexampleconcernsthelong-standingsoftwoodtimbertradedisputebe-tweenCanadaandtheUnitedStates.BecauseamajorityoftimberinCanadaispubliclyowned,theU.S.lumberproducersalwayssuspectthattheirCanadiancounterpartsbene tfromagovernmentsubsidy.Inadisputetwoyearsago,theU.S.DepartmentofCommercedecidedtosetadutyof27.22%onCanadiansoft-woodlumberimports.Sincethennegotiatorsonbothsideshavediscussedvarioussolutionsinordertoliftthistax.InthecaseofBritishColumbia(BC),asolutionagreeduponistoincreasethenumberofauctionsofpublictimber(i.e.,auctiono cuttingrights)andusetheauctionoutcomestopricetimberheldunderlong-termtenurelicenses.2Aquestionofinterestistheextenttowhichthepricesresultingfromtheproposedauctionsalesarereliablesignalstoallocatecuttingrightsontherestofpubliclands.Furthermore,doesitmatterthatdi erentauctionformatsmaybeused?Andwillbiddershaveincentivestocolludeinthetimberauctions?

Google’sIPOandBC’stimberauctionsarejusttwoofmanyexamplesintoday’seconomythatposeinterestingquestionsandchallengestoindustrialeconomists.Recentdevelopmentsinthe eldmayshedsomelightsonGoogle’sbusinessmodel,itsIPOauctiondesign,BC’stimberauctiondesign,andmoregenerallyonvariousindustrialpolicies.ToincreaseourunderstandingofbusinessstrategiesandindustrialpolicyissuesinCanada,thispaperattemptstoprovideaselectivereviewoftheindustrialeconomicsliterature.Forearliersurveysoftheliterature,wereferthereadertoSchmalensee(1988),Tirole(1988),andthechaptersintheHandbookofIndustrialOrganizationeditedbySchmalenseeandWillig(1989).Fisher(1989)andShapiro(1989)alsoprovideinterestingdebatesandthoughtfuldiscussionsontheliterature.

Thispaperisorganizedasfollows.Inthenextsectionwediscussbrie ysomeoftheimportantmethodologicalcontributionsofgametheorytotheanalysisofbusiness

FormoreinformationonthetradedisputebetweenthetwocountriesandonCanadiansoftwoodtimberauctions,seethearticle“TheSoftwood-lumberDispute,ASimpleLessoninEconomics,”TheEconomist,January30th,2003,andarecentpaperbyPriceandList(2004).2

strategies,publicpoliciesandtheperformanceofmarketmechanisms.Section3reviewstheburgeoningliteratureontheeconomicsofnetworkindustries,includingairlinenetworks,accesspricingandinterconnectionpoliciesintelecommunications,creditcardpaymentsystems,industryagglomeration,aswellastheroleofnetworkexternalitiesandthetheoryoftwo-sidedmarkets.Section4isdevotedtothetheoryofmarketdesignwithafocusonauctiondesign.Inthelastsectionweconsiderforeigndirectinvestment,atopicattheintersectionofindustrialeconomicsandinternationaltradetheory.

2.

2.1.GameTheoryandIndustrialEconomicsGame-theoreticModelsofBusinessStrategiesandPublicPolicies.Gametheoryprovidesaframeworkforanalyzingbusinessstrategiesandgovernmentpolicies.Byformallyspecifyingtheplayers’availableinformationandstrategies,aswellasthesequenceofchoices,gametheoryhelpspredicthowplayersbehavestrategicallyandevaluatetheconsequencesofthestrategicinteractions.Insteadofprovidingacompletereviewofthetheoryweuseanumberofexamplestoillustratetheusefulnessofgametheoryinmodelingcompetitionandbusinessstrategies.

Webeginwithabriefdiscussiononsimultaneous-movegames,orsometimescalledgamesinnormalform.Itconsistsofthreebasicelements:Anumberofplayers,asetofstrategiesforeachplayer,andapayo functionforeachplayerwhichmapsthewholestrategypro leintoherpayo s.Thewell-knownCournotmodelandBertrandmodelofoligopolyaregoodexamplesofsimultaneous-movegames.Thegame-theoreticformulationofoligopolymodelsisnotlimitedtotheanalysisoftheclassicalCournotorBertrandmodels.Theanalysiscanbeextendedtoavarietyofstrategiesthat rmsmayuse,includingcapacitydecision,qualitychoices,productdi erentiation,R&Dinvestment,advertising,andsoon.3

ItiswellknownthattheuniqueNashequilibriuminthesymmetricBertrandmodelissuchthatboth rmssetpricesequaltotheirmarginalcostandtheydonotmakeanypositivepro t.ThisresultisknownastheBertrandparadox,“Oneismonopoly,twoisperfectcompetition”.Itwouldseemtocontradictcommonob-servationsintwoways:inmarketswithfewsellers, rmstypicallydonotsellatmarginalcosts;andeveninperiodsoftechnologicalanddemandstability,pricesinoligopolisticmarketsarenotalwaysstable.GametheoryishelpfulinresolvingtheBertrandparadoxbyexamining,forexample,theroleofcapacityinpricecom-petition.Inpractice, rmstypicallydonothaveenoughcapacitytoservicetheentiremarketdemandwhenundercuttingpricestowardtheirmarginalcosts.Inthe3SeeShapiro(1989)forsomeexamplesandgeneraldiscussions.

shortrun,capacityis xedandcustomersarehencerationed.Thisimpliesthatthe rmwithahigherpricewillfaceapositiveresidualdemand,leadingtoequilibriumpricesabovemarginalcost.However,ifcapacityitselfisviewedasastrategicvari-able, rmswillanticipatetheimpactofsaidcapacitychoicesonsubsequentpricecompetition.Iftheyanticipatecorrectly,howmuchcapacitywillduopolistschoose?Game-theoretictoolsofextensive-formgames(orsequential-movegames)andtheprincipleofsubgameperfectionhelpprovideananswer,whichwewilldiscussbelow.

Commitmentisimportantinbusinessstrategy.Beingabletocommitearlyoftenleadstoa rst-moveradvantage,whichisparticularlysigni cantinthemoderneconomywherenetworkexternalityisprevalent.Thee ectivenessofcommitment,however,dependsonwhethersuchanactioniscredible.Withoutunderstandingtheroleofcredibility,anyrecommendationofbusinessstrategiesandgovernmentpolicieswouldbeine ective,andsometimesincorrect.Toanalyzeissuesrelatedtocommitmentandcredibility,weneedtomodelsequencesofmovesinagame.Thisiswhatthetheoryofgamesinextensive-form(orsequential-movegames)o ers.Toruleoutnon-crediblethreatsanotherequilibriumconceptisoftenused,namelysubgameperfectequilibrium.Subgameperfectequilibriacanbecalculatedbysolvingtheextensive-formgamebackward.

Oneofthemostimportantgamesinextensive-forminindustrialeconomicsispresentedinthefollowingexample.Dooligopolistschoosepricesorquantities?KrepsandScheinkman(1983)proposeaningenuousanswertothisquestion.Theystudyatwo-stagegameinwhich rms rstsimultaneouslychoosetheircapacityandthensimultaneouslyselectpricesafterobservingthecapacityofall rms.Consideronlytwo rms.Apurestrategyfora rmconsistsofacapacitylevelandapricefunctionofcapacitypair.Thepayo speci cationdependsonwhichrationingruletouse,i.e.,whathappenswhenthelow-price rmdoesnothavethecapacitytoservetheentiremarket.Underthee cientrationingrule,consumerswiththehighervaluationsbuyatthelowerprice.Undertheproportionalrationingrule,everyconsumerhasthesameprobabilityofbuyingatthelowerprice.Assumingthee cientrationingruleandconcavedemandfunction,KrepsandScheinkman(1983)showthatatthesubgameperfectequilibriumtheduopolistschoosecapacitylevelsequaltotheCournotquantitylevels.Inotherwords,the rmsbehavelikeCournotcompetitorsinthe rststageandlikeBertrandcompetitorsinthesecondstage.Theycommitnottoengageinruthlesspricecompetition,andasaresult,theequilibriumofthepricinggamewillbeoneinwhicheach rmsetsthesameprice.ThatpriceturnsouttobeequaltothepriceresultingfromtheCournotmodel.

ThisexampleprovidesalinkbetweentheBertrandpricecompetitionandCournotquantitycompetition.Moreimportantly,asarguedinTirole(1988),thistwo-stage

gameapproachillustratesageneralinsightthat rmsmaybeabletousenon-pricestrategiestosoftenpricecompetition.IntheKrepsandSheinkmanmodel,each rmrestrictsitscapacityasacommitmentnottochoosealowpricelater.An-othernon-pricestrategycanbeanychoiceofscalethatdeterminesthe rms’costfunctionswhichwouldinturndeterminetheconditionsofpricecompetitioninthesubsequentstage.Othernon-pricestrategiesusedtosoftenpricecompetitionmayincludeproductdi erentiation,most-favoredcustomerclauses,divestiturestrategies(TanandYuan,2003),frequent- yerprogramsinthecontextofairlines(BanerjeeandSummers,1987)andcustomerloyaltyprogramsinotherindustries.

Sometimes rmsmaymakeanearlycommitmentinordertostealcustomersfromtheircompetitorsorputtheircompetitorsinadisadvantageouspositionbyalteringtheircostfunctions.Therearestrategiesthat rmsoftenusetoraiserivals’costsorreducerivals’demands,someofwhichhavesigni cantcompetitionpolicyimplications.

Anotherimportantapplicationofsubgameperfectioninextensive-formgamesistocheckthecredibilityofentry-deterrencestrategies.ThisisnicelyillustratedinapaperbyJudd(1985).Theliteratureonentryhasarguedthatearlyinvestmentbyanincumbent rmtooccupymanyrelatedmarketsisane ectiveentry-deterrencestrategy.Judd(1985)providesaformalmodeltoarguethatastrategyconsistingincrowdingthemarketmaynotbecredibleandthereforenote ectiveinmanycontexts.Whenamulti-productincumbent rmcompeteswithanentrantinoneofitsmarkets,demandinmarketsforsubstitutegoodsfalls.If xedcostsarenotcompletelysunkandexitcostsarelow,theincumbentmaybebettero bywithdrawingitsproductandconcedingthemarkettotheentrant.Anticipatingthisoutcomeofproductmarketcompetition(backwardinduction),apotentialentrantwouldenter.Thus,thepotentialentrantisnotdeterredbytheincumbent rm’sinvestmentinamarket.

Inarelatedpaper,Hendricks,PiccioneandTan(1997)obtaintheoppositecon-clusionwhenproductsarecomplementsinsteadofsubstitutes.Theyshowthatamonopoly rmthatoperatesinseveralmarketswithcomplementarygoodshasstrongincentivestopricelowonthemarketwherethereisanentrantsinceother-wiseitspro tsintheothermarketswouldbereduced.Whenthecomplementaryrelationisstrongenoughorthenumberofcomplementaryproductslargeenough,theoptimalresponseofanincumbenttoentryisnottowithdrawitsproduct,evenifthenentrantstays.Thisanalysissuggeststhatcomplementaritiesamonggoodscanbeabarriertoentry.

Bothexamplesillustratethattheimportanceofbackwardinductioninanalyzingthecredibilityofbusinessstrategies.Anotherclassofgamesinextensive-formthathasmanyapplicationsinindustrialeconomicsisrepeatedgamesorsupergames.By

replicatingaBertrandgameorCournotgame nitelyorin nitelymanytimes,wecanapplythebackwardinductionmethodandsubgameperfectequilibriumconcepttostudytheimportantissueoftacitcollusion.

Twomoreapplicationsofgametheoryinmodellingstrategicbehaviorwarrantsomediscussion:Strategictradepolicyandmanagerialincentives.Anationalgov-ernmentmayhaveincentivestosettradepolicieshelpingits rmswhentheycompeteagainstforeign rmsinaninternationalmarket.Thesepoliciesmayincludedirectsubsidies,taxbene ts,andlow-interestloans.Suchpoliciesoftenimprovedomestic rms’strategicpositionsandputrivalsinadisadvantageousposition.However,ifallgovernmentsusesimilarpolicies,theymaynotnecessarilybene t.Thisprob-lem,whichinspiredalargeliteratureonstrategictradepolicy,was rststudiedbyBranderandSpencer(1985).Similarargumentshavebeenusedtostudymanage-rialincentivesbyFershtmanandJudd(1984).Whenmanagersplayanoligopolygame,ownersmaywanttomotivatemanagersbycompensatingthembasedonacombinationofpro tsandrevenues.

Firmsoftenhaveprivateinformationabouttheircostsandmarketdemand.Toanalyzetheimpactofprivateinformationon rms’strategies,weneedtoapplythetheoryofgameswithincompleteinformation.AgoodapplicationistheBertrandpricecompetitionmodelwithprivatelyknownmarginalcosts.Supposethatanum-berof rmscompetebysettingpricessimultaneously.Each rmknowsitsown(constant)marginalcostbutcannotobserveitsrivals’marginalcosts.Assumethatthemarginalcostsareindependentlydrawnfromthesamedistribution.Each rm’sstrategyisafunctionofitsownmarginalcost.The rmsbehaveasBayes-Nashplayerschoosingpricesbasedontheexpectationsoftheother rms’equilibriumstrategies.Inasettingwithgeneralcostfunctions,Spulber(1995)showsthatthereexistsauniquesymmetricequilibriumpricingstrategy,andthatinequilibriumall rmsexceptforthehighestcosttypemakepositivepro ts.ThisanalysisprovidesanalternativewayofresolvingtheBertrandparadox.

Moreimportantly,thisexampleprovidesalinkbetweenoligopolytheoryandauctiontheory.Thesettingisessentiallythesameasthestandardauctionenviron-mentwithindependentprivatevaluesandvariableunitsofquantities.TheBertrandcompetitionisliketheruleof rst-pricesealed-bidauctions.ABayes-Bertrand-Nashequilibriumisapro leofbiddingstrategiesthatcanbecomputedusingthestandardtechniqueinauctiontheoryandgametheory.

Manyprocurementprojectswith xed-pricecontractsarecompetitivelyawardedthroughBertrand-like rst-priceauctionprocedures.Examplesincludehighwaycon-structioncontracts,schoolmilkdeliverycontracts,anddefenseprocurementcon-tracts.AswewilldiscussinSection4,thetheoryofmarketdesigndealswith

designingtherulesofgameswithincompleteinformationplayedbetweencompeting rms.

Acombinationofgamesinextensive-formandgamesofincompleteinformationleadstodynamicgamesofincompleteinformation.Thisclassofgamesalsohasmanyapplicationsinindustrialeconomics.TheperfectBayesianequilibriumandthesequentialequilibriumconceptsareusedtoanalyzeimportantissuessuchasreputationbuilding,sequentialbargainingwithincompleteinformation,andpriceorwarrantyasasignalofquality.

Intherestofthissectionwefurtherdevelopafewexamplestoanalyzestrategicbehaviorandtheexerciseofmarketpower.

plements,SubstitutesandStrategicBehavior.Supposethemar-ketdemandsfortwoproductsareinterdependent.Thetwodemandfunctionsmaynotbesymmetricsincetheconsumersforoneproductmaygenerallydi erfromthoseoftheotherproduct.Forexample,televisionprogrammeviewers(ornewspa-persubscribers)andadvertisersaretwodistinctgroupsofconsumers;theaggregatedemandsforTVprogrammes(ornewspapers)andadvertisementsareinterdependentbutnotnecessarilysymmetric.

Whenthetwoproductsaredemandsubstitutes,thereisapositiveexternalitybetweenthetwopricesinthesensethatanincreaseinthepriceforoneproductin-creasesthedemandfor,aswellastheseller’revenuefrom,theotherproduct(giventheconstantmarginalcost).Whilecompetitioninthemarketfordi erentiatedsub-stitutegoodstendstoreduceprices,amulti-productmonopolistcaninternalizethepositiveexternality,leadingtopro t-maximizingpriceshigherthanthelevelsthattwoindependentdivisionsorcompeting rmswouldchoose.Ontheotherhand,whentheproductsaredemandcomplementary,thereisanegativeexternalitybe-tweenthetwopricevariableswhenevaluatingthepro t.Inthiscase,themonopolistchargeslowerpricesthantwoindependentdivisionsor rmswoulddo.Thisout-comeistraditionallyemphasizedinthecontextofaverticalmanufacturer-retailerrelationshipwheredoublemarginalizationcanbeeliminatedbyverticalintegrationorsomeformsofverticalrestraints.

Thenegativeexternalityarisesalsoinmanyothercontextswhereademandcomplementarityispresent.Inthosesituations,the rmsmaynotcompeteinthetraditionalsense.Cooperationcanactuallyhelpreduceconsumerpricesandincrease rmpro ts.CooperationinsuchacontextisemphasizedinabookentitledCo-opetitionbyBrandenburgerandNalebu (1996).Theyusethewordcomplementortodescribetherelationshipbetween rmsthatprovidecomplementaryproductsandarguethat rmsaregenerallycomplementorsinmakingmarketsandcompetitors

individingupmarkets.Recognizingdemandcomplementarityandtheresultingnegativeexternalityinpro tfunctionshasimportantimplicationstounderstandinganddesigningbusinessstrategies.

Forinstance,inthecontextofmulti-productmonopoly,thenegativedemandex-ternalityoftenimpliescross-productsubsidization.Examplesofcross-productexter-nalitiesincludeAbobeAcrobatreaderandwriter,newspaper(orTV)subscriptionandadvertisingservices,telephone/cellularphoneservices(callersandreceivers),shoppingmallswithdi erentretailers(aswellasretailersandcustomers),creditcardissuersandacquirers,andsoon.Thesemarketsaresometimescalledtwo-sidedormulti-sidedmarketssincetwoormoregroupsofparticipantsneedtointeractvia(monopolisticorcompetitive)intermediaries.WewilldiscussthisissueinSection

3.6.

Inacompetitivesituation,integrated rmsmaybeabletousetechnologicalinstrumentstocontrolthedirectionofcross-productexternalitiesandputnon-integratedrivalsinadisadvantageousposition.Anearlyexampleconcernsadisplaybiasassociatedwithcomputerreservationsystems(CRS).EarlyinthedevelopmentofCRS,theairlinesthatownedthemhadincentivestolisttheirown ightsbeforetheirrivals’ ights.Duetoastandardagencyproblem,travelagents,knowingtheexistenceofthebias,maynothaveincentivestospendtimetosearchforthebestschedulefortheircustomersandwouldeasilychoosethescheduleonthetopoftheircomputerscreen.FisherandNeels(1997)estimatethatduetothisdisplaybias,$58millionwasdivertedbytworeservationsystems(i.e.,SabresystemownedbyAmericanAirlinesandApollosystembyUnitedAirlines)in1984.Eventually,theCivilAeronauticBoardintheU.S.forbadesuchpractice.Inthisexample,therearethreecomplementaryproducts:travelreservationservices,reservationsystems,andairtravels.Travelagentso ertheircustomeraccessto ightinformationprovidedbycomputerreservationsystemsandorganized,indirectly,byairlines.

Internetmarketsshareasimilarfeaturewiththecomputerreservationsystem:informationshowingonthetopofthecomputerscreenoronthe rstscreenofanInternetportaltendstobemorevaluable.LuandTan(2003)provideamodeltoanalyzehowthe rstscreenadvantageandthestickinessofwebsitescanhelpunder-standmanyrecentmergersbetweenInternetserviceprovidersandcontentproviders.Aserviceproviderhastheadvantageofcontrollingthe rstscreenseenbyitssub-scribersandisabletocreatesomeformofstickinesstoprolongeachsubscriber’sstayatitswebsite.Thestickierisawebsite,themoreadvertisementsitwillgenerate.Thecontentproviderthenhasincentivestointegratewithitsserviceproviders.Acombinationofthe rstscreenadvantageandthestickinessofwebsiteso ersincen-tivesfor rmsincomplementarymarketstoexploitthedemandcomplementarity

acrossmarkets.

Complementarityissometimesinducedbyaproductiontechnologythatexhibitseconomiesofscale.Intheairlineindustry,forinstance,strongeconomiesofdensitygiverisetohub-spokenetworkswhichinturninducecomplementaryservicesbetweenconnecting ights.Amonopoly rmthatoperatesahub-spokenetworkhasincentivestopricelowontheconnectingmarketwherethereisanentrant(Hendricks,Piccione,andTan,1997).Therefore,demandcomplementarityacrossproductscanbeabarriertoentryaswell.

Moreover,whenbothcomplementaryandsubstituteproductsarepresent,twoopposingtypesofdemandexternalitiesarise.Firmsmay nditoptimaltodesignstrategiestoo settheopposingexternalities.ThisisillustratedinTanandYuan(2003),wheretwocompetingconglomerateshaveincentivestodivesttheircomple-mentaryproductlines.

2.3.BundlingandTied-inSale.Thepracticeofgroupingseveralproductstogetherandsellingatonepriceisreferredtoasbundling.Bundlingstrategyispervasiveinpractice.Economistscallontwodominantexplanationsforbundling:pricediscriminationandentrydeterrence.4Theideaofpricediscriminationis rstillustratedinStigler(1963)’sclassicalexampleofcinemablockbookingschemesinwhichatheaterisrequiredtotakeapackageofdi erentmoviesinsteadofonlysinglereleases.AfurtheranalysisispresentedbyAdamsandYellen(1976).Accordingtothem,whenconsumers’valuationsformultipleproductsare(perfectly)negativelycorrelated,bundlingenablesthemonopoly rmtoreduceheterogeneityinvaluationsandhencecapturemoreconsumersurplus.AgeneralformulationisprovidedbyMcAfee,McMillanandWhinston(1989).Theyshowthatthemonopolistprefersmixedbundlingtonobundlingevenifconsumers’valuationsareindependent.Whenvaluationsarecorrelated,theyalsoprovidesu cientconditionsformixedbundlingtodominatenobundling.

Tying(ortied-insale)isaspecialformofbundling.Itreferstoconditioningthesaleofonegoodonthepurchaseofanother.Thisoftenoccurswhenthe rstproductismonopolisticallysuppliedwhilethesecondproductmarketiscompetitive.Thecentralquestioniswhetherthemulti-productmonopolistisabletoextenditsmonopolypowerinthe rstmarkettothecompetitivemarket.Theliteratureo erstwooppositeviews.Theleveragetheorysays“yes”whiletheChicagoSchoolassertstheopposite.Itturnsoutthattheanswerdependonthenatureofconsumers’

Anotherexplanationforbundlingissimplycostsavings.Notethatmostproductsareabundleofseveralcomponents.Insteadoflettingconsumersputdi erentcomponentstogether,itisoftencost-e ectiveforthesellertoo erabundledproduct.SeeSalinger(1995)foradiscussion.4

demands.

Inthecasesofunitdemandsorperfectcomplements,themonopoly rmneednottyinginordertoreachthemonopolyoutcomeonbothmarkets.Ontheotherhand,ifthedemandsforthetwoproductsareindependentanddownward-sloping,andifthemonopolistisrestrictedtouseuniform(orlinear)pricing,tyingispro table.Thebasicinsightisthatifthemonopolistisunabletoextractfullytheconsumer’ssurplusinthemonopolymarket(e.g.,duetothelimitationoflinearpricing),tyingprovidesaninstrument(ormeteringdevice)forthe rmtodiscriminateacrossconsumersandextractmoreconsumersurplus.Inadeterministicenvironmentwithvariabledemandsandlinearpricing,MathewsonandWinter(1997)provideanecessaryandsu cientconditionfortyingtobepro table.Inparticular,theyshowthattyingisnotpro tableforalllevelsofconstantmarginalcostsifandonlyifthetwogoodsarestronglycomplementary,i.e.,theratioofthedemandsforthetwoproductsisindependentofoneoftheprices.Thisconditionissatis edwithperfectcomplementsbutviolatedwithindependentdemands.Their ndingsummarizespreciselythepreviousliteratureontying.

Ifthepro tabilityoftyingisduetothemonopolist’sinabilitytoextractfullytheconsumer’ssurplusinthemonopolymarketbyusinglinearpricing,thequestionisthenwhethertyingisusefulwhenthe rmcanimplementnon-linearpricingsuchastwo-parttari s.ThesecondpartoftheMathewsonandWinter(1997)paperaddressesthisquestion.Intheirsetting,demandsareuncertainandarerepresentedbyarepresentativeconsumer’s(quasi-linear)indirectutilitywithtwoparametersthatmeasuretherelativewillingnesstopayforthesetwoproductsandthatarenotobservedbythe rm,butthe rmknowstheirjointdistribution.5Ifthemonopolistcanimplementtwo-parttari s,MathewsonandWintershowthattyingispro tablewhenthetwoparametersarepositivelycorrelated.Thatis,evenifthe rmcanusetwo-parttari stoextractmoreconsumersurplusthanlinearpricingdoes,tyingcanfurtherimprovepro tability.Notethattwo-parttari smaynotbethebestnon-linearpricingschemeforthemonopolistwhentyingisnotallowed.Thequestionstillremainsifthemonopoly rmprefersthesecond-bestnon-linearpricinginthemonopolymarkettotying.

Whiletheearlyliteratureontyingmostlyassumethatthesecondproductmarketisperfectlycompetitive,Whinston(1990)isthe rsttorelaxthisassumption.Inhismodel,themonopoly rmandonerivalo erdi erentiatedproductsinthesecond

Theadvantageofusingthequasi-linearutilityfunctionisthatthereisnoincomee ect.Thisisoftenusedinapplicationssinceonecanconvenientlycalculateconsumers’surplusandconductpar-tialequilibriumwelfareanalysis.Inarecentpaper,Duhamel(2004)providesageneralequilibriumfoundationforarbitrarypartialequilibriumwelfareanalysisinsecond-besteconomies.5

market.Ifthemonopolistcandecidewhethertobundlebeforethetwo rmscompeteinprices,Whinstonshowsthattyingmakesthemonopolistmoreaggressiveanddiscouragesentry.However,iftherivalstaysinthemarket,bundlingmayactuallyhurtthemonopolist.Credibilityofcommittingtobundlingearlyiscriticalinthismodel.

CarltonandWaldman(2002)consideranothersettinginwhichanentrantcanprovideacomplementaryproductthatissuperiortotheincumbent’ssecondproduct.Iftheentrantcommitstoenteringthecomplementaryproductmarketonly,itsentryincreasesindustrypro t,whichinturnmaybene ttheincumbent rm.However,iftheentrantmightlaterentertheoriginalproductmarket,theincumbent’spro twouldbea ected.Inthiscase,theincumbenthasincentivestoo erbundlingofthecomplementaryproductsinitiallywhichcouldnotbematchedbytheentrantbecauseofitssmallscale.Therefore,tyingcanbeusedtodenyasmall-scaleentrantanddeterentry.

ArecentpaperbyNalebu (2004)o ersanotherperspectiveontying.Inhismodel,theincumbentisatwo-productmonopolistfacingapotentialentrantthatmayenterineithermarket.Heshowsthatbundlingallowstheincumbenttouseitsmonopolypowerineachmarkettoprotectitsstatusintheothermarket.Sincebundlingisalsopro tableabsententry,itisacredibletoolforentrydeterrence.Nalebu furthershowsthatinhismodelthegainsfromthepricediscriminatione ectaregenerallysmallascomparedwiththegainsfromtheentry-deterrencee ect.

2.4.CountervailingBuyerPower.Marketpowerofdominantmanufacturersmaybeconstrainedbyindependent,dominantretailersorbuyers.Theroleofbuyerpowerhas rstbeenaddressedinanearlybookentitledAmericanCapitalism:TheCountervailingPowerbyGalbraith(1952).Hearguesthatlargeretailorganizationssuchasthemajorchainstoreoperatorsareabletoexercisecountervailingpowerovertheirsupplierstolowerwholesalepricesandarewillingtopassthesesavingstotheircustomers.Thisleadshimtoconcludethatretailers’countervailingpowerissociallydesirable.However,Galbraithdoesnotexplainwhyretailerswouldhaveincentivestopasscost-savingstoconsumers.

Inarecentpaper,Chen(2003)providesaformalmodeltoexaminethiscountervailing-powerhypothesis.Chenshowsthatanincreaseintheamountofcountervailingpowerpossessedbyadominantretailercanleadtoafallinretailpriceforconsumers.Healsoillustratesthatduetopossiblee ciencyloss,totalsurplusdoesnotalwaysin-creasewiththeriseofcountervailingbuyerpower.Chenfurtherarguesthatthepresenceoffringecompetitioncanbecrucialforcountervailingpowertobene tcon-sumers.Thisanalysishasimportantimplicationsincompetitionpolicyenforcement.

TworelatedpapersontheroleofcountervailingpowerarevonUngern-Sternberg(1996)andDobsonandWaterson(1997).They ndthatincreasedconcentrationattheretaillevelcanleadtohigherpricesforconsumers.AsChen(2003)argued,theirmodelsdonotcapturesomeoftheimportantfeaturesoftheretailindustry.

workIndustries

Manynetworkindustrieshavebeenundergoingrapidtechnologicalandregulatorychanges.Somehavebeentransformed,orwentthroughatransitionfromtradition-allyregulatedtocompetitivesectors.Examplesincludetransportationindustriessuchasairlines,railroad,truckingandexpresspackagedelivery,publicutilityin-dustrieslikeelectricity,aswellastelecommunicationsandcableTVindustries.TheInterneto ersanotherexampleinwhichnetworkfeaturesareobviouslysigni cant.

Networkindustriescanbequali edwithanumberofcharacteristics.The rstissigni canteconomiesofscaleinproductionforwhichairlineindustryisanex-ample.High xedcostsandrelativelylowmarginalcostattheroutelevelgiverisetoeconomiesofdensities,whichinturninducesaparticularnetworkpatternsuchashub-spokenetworks.Theproductionofinformationproductssuchascomputersoftware,recordedmusicandmoviesofteninvolveshigh xedcostandlowmarginalcostaswell.Thesecondcharacteristicconcernscomplementarityamongproductsandservices.Thecomplementaritybetweenhardwareproducts(computers,DVDplayers,andvideogameplayers)andsoftwareproductsisagoodexample.Printme-dia,TVbroadcastingandInternetportalsallo ercomplementaryservicestoviewersandadvertisers.Thethirdcharacteristicofnetworkindustriesis(directorindirect)networkexternality,plemen-tarityandnetworkexternalitiesofteninducemarketstobetwo-sidedormulti-sided.Thecreditcardsindustryisonesuchanexample.

Businessstrategiesandregulatorypoliciesinnetworkindustrieshavereceivedmuchattentionfromresearchersinlate80sand90s.Inthissection,weselectivelyreviewsomeofthesestudies.AmorecompleteandcomprehensivediscussiononnetworkindustriesisprovidedbyShy(2001)inhisbookTheEconomicsofNetworkIndustries.

3.1.Hub-SpokeNetworks.Amongallthetransportationsectors,theairlineindustrypresentsaninterestingcase,duetoitsnetworkformation.Followingthederegulationofthesectorinthelate70s,manynew rmsentered,butsincethentheindustryhasgrownincreasinglymoreconcentrated.Giventhefreedomtoenter,exit,andchoosetheirownroutestructureandairfares,mostairline rmshavetransformedtheirnetworksintohub-and-spokesystems.Astrikingfeatureofthesenetworksis

thatairline rmsusedi erentcitiesastheirhub-cityinsteadofsharingthesamehub-city.

Thereisanextensiveempiricalliteratureontheairlineindustrywhichtriestomeasurethee ectofnetworkcharacteristicsandnumberofcompetitorsonprices6However,acloserlookattheindustrysuggeststhatalongwithprices,networkchoicesoughttobeendogenouslydetermined.Inthetheoreticalliteratureonairlinenetworks,somemodelshavebeendeveloped.BruecknerandSpiller(1991,1994)con-sideralinearmodelinwhicheach rmchoosesquantitiesgivenaparticularnetwork.Oum,Zhang,andZhang(1995)examinethee ectsofthestrategicinteractionbe-tweenderegulatedairlinesontheirnetworkchoices.Inparticular,theyaskwhetherswitchingfromalineartoahub-spokenetworkcanbeexplainedby rms’pursuitofstrategicadvantages.Zhang(1996)identi esanegativeexternalityoflocalcompe-titioninairlinehub-spokenetworks.

Inaseriesofarticles,Hendricks,PiccioneandTan(1995,1997,1999)provideageneraltheoreticframeworktoanalyzenetworkchoices.Theyconsideranenviron-mentwheretherearencitiesandindividualsineachcitywhowishtotraveltotheothercities.Thereareseveral rmsintheindustry,eachofwhichhavingtochooseanetworkof ightsandasetofpricesfortravellinginthatnetwork,tomaximizeitspro tsgiventhechoicesoftheother rms.Thus,bothnetworkdesignandpricesareendogenous,andthemarketpricesandnetworksthatresultrepresentanequilibriumoutcome.Theycharacterizethisoutcomeinavarietyofdecisionenvironments.

Hendricks,PiccioneandTan(1995)focusontheoptimalnetworkandpricesforanunregulatedmonopoly rmwhendemandsandcostsaresymmetric.Theirmainresultisthat,ifthereareeconomiesofdensityinthenumberofindividualstravellingbetweentwodirectlyconnectedcities,theoptimalnetworkiseitherahub-spokenetworkorapoint-to-pointnetworkinwhicheverypairofcitiesisconnecteddirectly.Thehub-spokenetworkpossesseshighertra cdensitiesthananyothernetworkswithmoredirectconnections,permitstheuseoflargerplanes,andconsequentlylowersthetotalcostsofsatisfyingagivensetoftraveldemands.Therefore,thecoststructureisonedrivingforcefortheemergenceofhub-spokenetworks.

Hendricks,PiccioneandTan(1997)o eranexplanationastowhymajorhubairportsaredominatedbyafewaircarriersasindicatedbyanumberofempiricalstudies.Inahub-spokenetworkwithncities,connecting ightsarecomplementarygoods,petitionbetweenaregionalandanationalcarrierononeofthese

ExamplesincludeBorenstein(1989),ReissandSpiller(1989),MorrisonandWinston(1990),Brueckner,DyerandSpiller(1992),BruecknerandSpiller(1991,1994),Berry(1992),Kahn(1993)andBambergerandCarlton(1993).6

segmentslowerspricesandpro tsonthatsegment.However,thenationalcarriercano setsomeofthelossesbyadjustingpricesonthen 2complementarysegmentswheredemandshaveincreased.Whenthesizeofthenetworkislargeenough,thehuboperator’soptimalresponsetoentryinaspokemarketisnottowithdrawits ightsfromthatmarket,eveniftheregionalcarrierstays.Asaresult,regionalcarriersareforcedtoexitandentryisdeterred.Thus,ahub-spokenetworkisacredibletoolfordeterringentry.

Empiricalevidenceillustratesthatairlinesusuallyoperateseparatehub-spokenetworksandcompetehead-to-headforthetra cbetweennon-hubcities,buthavealocalmonopolyonthetra ctoandfromtheirrespectivehubcities.Thisraisesanumberofinterestingquestions.Whydo rmschoosedi erenthubcitiesinsteadofsharingthesamehubcity?Whydoesasingle rmoperatemorethanonehub-spokenetwork?Doescompetitiongeneratetoomanyhubs?Domergersrepresentattemptsby rmstoexploiteconomiesofdensities?Ifso,howmany rmscantheindustrysupport?Toanswerthesequestions,theentireairlinenetworkneedstobetreatedasanequilibriumoutcome.

Hendricks,PiccioneandTan(1999)investigatetheconditionsunderwhichhub-spokenetworksareanequilibriumcon gurationinaduopoly.Inthemodeltwocarrierschoosenetworkstoconnectcitiesandcompeteforcustomers.Theyshowthatifcarrierscompeteforcustomersaggressively(e.g.,Bertrand-likebehavior),onecarrieroperatingasinglehub-spokenetworkisanequilibriumoutcome.Ifcarriersdonotcompeteaggressively(e.g.,implicitcollusionviafrequent yerprograms,andproductdi erentiation),anequilibriumwithcompetinghub-spokenetworksexistsprovidedthatthenumberofcitiesisnottoosmall.Theyalsoprovidesu cientconditionsunderwhichallequilibriaconsistofhub-spokenetworks.IntheU.S.,airlinesoperatenetworksthatappeartobepredominatelyhub-spoke.However,theycontainmorethanonehub-city,aswellassub-networksthatarepoint-to-pointoperations.Thismayre ectdistancefactorsthatleadcarrierstooperateregionalhub-spokenetworks,orotherfactorssuchasdemandasymmetriesandschedulingconstraints.FurtherresearchisneededtoexplorethetheoreticalimplicationsofrelaxingtheassumptionsoftheHendricks,Piccione,andTan(1999)model.

Anotherinterestingandimportantquestionisthatofinternationalairlineal-liancesthatareformedtotakeadvantageofnetworkcomplementarities.Becauseofnationalregulatoryrestrictionsacrosscountries,foreigncarrierscannotpickuppoint-to-pointpassengersondomesticroutes.Aninternationaltravelleroftenneedstocombineseveralsegmentsfromdi erentairlinestogether.Therearee ciencygainsfromcooperationamongtheairlinesbyinterconnectingtheirexistingnetworks,throughbettercoordinationoftheirconnecting ights,andbyjointadvertisingand

promotion.Buthowdodi erentalliancescompete?Andwhatdetermineswhichalliancetojoin?Iseconomicwelfareenhancedbystrategicalliances?Theseremainopenforfutureresearch.

Theanalysiscanbeextendedtostudyotherindustrieswhereeconomiesofnet-worksizearesigni cant.Forinstance,thenetworkpatternintheexpresspackagedeliveryindustryintheU.S.ismainlycharacterizedbyhub-spokenetworks.Theissueofinterconnectingnetworksownedbydi erent rmsalsoarisesinthisindustryaswellasinmail,Internet,andtelecommunicationindustries.

3.2.InterconnectioninTelecommunications.Countrieshaveregulationsthatlimittheoperationsofforeignoperators.Thisisparticularlytrueinthecaseofinternationaltelecommunications.Fixedandsunkcostsofbuildinganetwork(orinfrastructure)aresigni cant;andbuildingoverlappingnetworksisnotcost-e ectivefor rmsandnotsociallydesirable.Consequently,telecommunicationnetworksinmanycountriesnowadaysaremadeofamyriadofseparatelyownednetworks.Mostnetworksareinterconnectedtoformanetworkofnetworks,whichalsoenablescom-municationamongpeoplebelongingtodi erentnetworks.Technologicaladvancesandgovernmentderegulationpolicies,particularlynetworkaccessandinterconnec-tionpolicies,havemadeinterconnectionsofthesenetworksandaccessprovisionpossible.

Accesstoanothernetworkisacriticalinputtoo ercompetitiveservicesto nalconsumers.Theprovisionofaccessisalsoawaytosolvethenetworkexternality,whicharisesbecausecallersandreceiversneednotbewithinthesamenetwork.Bybuyingaccesstoanexistingnetwork,entrantsdonothavetobuildasimilarnetwork.Thisreducesentrybarriers.However,theaccesschargeisnotalwayspro-competitiveorneutral.Forexample,largenetworkoperatorsmayusehighaccesschargetoexcludesmallones;andsymmetricnetworkoperatorsmaysustainretailcollusionthroughanappropriatechoiceofaccesschargestoeachother.Mostoftherecentstudiesarecenteredaroundtheinteractionamongaccesscharge,marketstructureandretailcompetition.

SettlementRatesinInternationalTelecommunications.Historically,two-wayaccessstartedwiththeinternationaltelephoneservice.Theaccesscharge(perminute)inthisscenarioiscalledthesettlementrate.Itisgenerallydeterminedthroughcarriers’negotiationundervariousrestrictionssetbygovernments.Thebottleneck(terminatingcalls)ofacountry’snetworkusedtobeownedbyanationalmonopolisticcarrier,andisstillthecaseinmanycountries.CarterandWright(1994),amongothers,formallystudysettlementagreementsbetweentwonationalmonopolisticnetworks.Eachmonopolistprovidesinternationalcallservicesinone

countryandneedsterminationservicesbytheothercarrier.Thatis,themonopolistsprovidecomplementaryinputstoeachotherandthereisnodirectcompetitionoverservicesubscriptions.Ifthetwocarrierssetsettlementratesnon-cooperatively,theequilibriumratesarealwaysabovethemarginalcostofprovidingtheterminationservice.Duetoastandarddouble-marginalizatione ect,consumersinbothcoun-triespayhighretailpricesforinternationaltelephoneservices.Anexplicitcollusionoversettlementratesmayactuallyreducetheratestowardmarginalcostsofpro-vidingaccess,whichinturndecreaseconsumerpricesinbothcountries.Whenareciprocityofthesettlementratesisimposed,themarginal-costbasedsettlementratecanbereachedifthetwocountriesaresymmetric.

Inmanycountries,competitionhasbeenrecentlyintroducedinthelongdistancetelephonemarkets.Carrierscompetenotonlyinretailpricesforinternationalcallsbutalsointerminationservicesforforeigncarriers.Whencompetingcarriersnego-tiatesettlementrateswithamonopolisticcarrierintheothercountry,thewelfaregainfromcompetitioncanbeo setbyin atedsettlementrates.In1987,theU.S.FederalCommunicationsCommissions(FCC)initiatedasetofregulatorypolicies,calledInternationalSettlementsPolicy(ISP),toguidesettlementnegotiations.TheISPconsistsofthreemajorcomponents.1)Uniformity:AllU.S.carriersmustpaythesamesettlementratefortheoutboundtra conthesameroute.Thisruleisdesignedtomakecompetingcarriersbehaveasasingleentitywhennegotiatingoverthesettlementrates.2)Reciprocity:AllU.S.carriersmustreceivethesamerateforterminatinginboundtra cfromaforeignmonopolisticcarrierastheratepaidforoutboundtra c.3)ProportionalReturnRule(PRR):Tra cfromaforeigncarrierisallocatedamongtheU.S.carriersinexactproportiontotheirsharesofoutboundtra cfromtheU.S.tothatcountry.Thatis,PRRallocatesincomingtra cbycarriers’retailmarketshares.

Inarecentpaper,Ju(2004)studiestheimpactofISPonthedeterminationofretailpricesandsettlementrates.HeillustratesthatPRRexertsadownwardpressureonretailpricesbecausedomesticcarriersarecompetingfortheinitiationofout owtra candthesettlementofin owthroughonestrategicvariable.UnderISP,whennegotiatingthesettlementratewithaforeignmonopolyprovider,domesticcompetingcarrierswouldindeedhavetheincentivetopayahighsettlementrateinordertoalleviatetheretailcompetition,asthelossattheretailsegmentcanbecompensatedbyforeignsettlementpayments.Fromasocialperspective,settlementrateabovemarginalcostrepresentsasocialcosttotheeconomy.Alowrateisdesirableforthecompetitivecountry,asitusuallyhasnetout ow.Thiscallsforthegovernmenttofurtherrestrictthecarriers’negotiationbyimposingaratecap.The

BenchmarkPolicyadoptedbyFCCin1997isagoodexample.7

Analternativeallocationruleisa xed-divisionrulewherebycarriersdeterminetheexactallocationofincomingtra cpriortotheirchoicesofretailprices.A xed-divisionruleremovesthepricedistortioncausedbyPRR.Therefore,theindustrypro tishigherthanthatunderPRR.Galbi(1998)andRieck(2000)questionthedesirabilityofPRRwhenthein owislarge.Sincethesettlementofforeignin owbecomesmorepro table,carrierswouldsettheretailpricebelowthemarginalcost(includingthesettlementratepaidtoforeigncarriers)paringthewelfaregeneratedbyPRRwiththeonebya xed-divisionrule,Ju(2004) ndsthatgiventhesamesettlementrate,thesocialwelfareunderPRRishigherthanthatundertheequal-sharingrule.Healsoshowsthattheindustrypro tunderPRRislessvariabletothechangesofsettlementratesthanundera xed-divisionrule.ThisexplainswhytheimplementationoftheBenchmarkOrderreceivedlittleresistancefromtheU.S.carriers.

CollusioninLocalNetworkCompetition.Inthecaseoflocalnetworkcom-petition,interconnectednetworkproviderscompeteforthesamesetofsubscribers.Armstrong(1998)andLa ont,ReyandTirole(1998a,1998b,hereafterLRT)pro-videabenchmarkmodeltoanalyzethisindustrystructure.Themainconcernistheriskofcollusionthroughaccesscharges.

Intheirmodels,twosymmetricnetworks,eachlocatedatanextremeofalinearcity,competeforconsumersalaHotelling.Giventhepricescharged,eachconsumersubscribestooneandonlyonenetwork.Consumersbene tonlyfrommakingcalls.Withthesamecallingcharge,theinboundandoutboundcallsareassumedtobebalanced.8Whentheretailpriceisuniform(orlinearpricing)andtheaccesschargeisreciprocal,itisshownthat,ifthepriceequilibriumexists,accesschargecanbeaninstrumentfortacitcollusion.Thetwonetworkscancollectivelychooseasu cienthighaccesschargeforo -netcallstorecoverthemonopolyretailprice.Priceunder-cuttingincreasesmarketsharebutalsoincreasestheaccesspayment.Whentheaccesschargeishighenough,thegainfromretailiso setbytheaccesspayment,andanydeviationfromthemonopolypriceisnotpro table.

Whentheretailpricetaketheformofatwo-parttari ,i.e.,asubscriptionfeeandausagefeepercall,however,LRT ndaneutralityresultofaccesscharge.The xedfeeisasubstitutefortheaccesschargeandismoree ectiveatextractingsurplusthan

TheBenchmarkPolicyimposesacaponsettlementratesininternationaltelephoneagreements.TheratecapsaresetbyFCC.Thepolicyintendstoreducethesettlementratesthatarenegotiatedbythetelephonecarriers.

8Balancedcallingpatternimpliesanygivensubscriberisequallylikelytocallanyothersub-scriber,regardlessofthenetworkthatthereceiverison.7

theaccessfee,sincethe xedfeehasnoimpactondemand.Inequilibrium,theusagefeeistheperceivedmarginalcostofeachnetwork,andthesubscriptionfeeequalsthemarginalcostofaddingacustomerplusamarkupre ectingsubstitutability.Moreimportantly,theequilibriumpro tisindependentoftheaccesscharge.Thus,theaccesschargecreatesnoincentivetocolludeanditiseasytosetitatthesociallye cientlevel,i.e.,themarginalcostofprovidingaccess.

ThesecondpaperbyLRTallowspricediscriminationoveron-net(withinanet-work)ando -net(across-network)calls.Iftheaccesschargeisdi erentfromthemarginalcostofprovidingaccess,callsindi erentdirectionshavedi workoperatorshavetheincentivetochargedi erentlyoverthetwotypesofcalls.Ahigheraccesschargeleadstoahighero -netprice,butalsoindi-rectlydecreasestheon-netprice.The nale ectofanincreaseinaccesschargewouldbealoweraveragepriceandpro t.Asnotedbytheauthors,“thewedgebetweenon-netando -netpricesisdetrimentaltoconsumptione ciency,butmayintensifycompetition,withambiguouswelfaree ects.”Highaccesschargesarethereforenotnecessarilyacollusivedevice.9

3.3.CreditCardPaymentSystems.Thecreditcardpaymentindustryex-hibitsstrongdemandcomplementarity.Atypicalcreditcardpaymenttransactioninvolvesfourparties:aconsumerandamerchant,eachdealingwitha nancialinstitution.Themerchant’s nancialinstitution(mostlybanks)isreferredtoas“acquirer”andprovidesafacilitytoperformthetransaction.Theconsumer’sbankisthe“issuer”andendowstheconsumerwithacardtoinitiatethesetransactions.Theissuerandacquirermustcooperatetoprocessatransactionandexchangeaninterchangefeewhendoingso.ThedominantcreditcardsystemsareVisaandMasterCard,eachofwhichisanassociationofissuersandacquirerswithopenmem-bership.10Theassociationshavedesignedasetofrulestogoverntheinteractionamongmemberbanksandparticularlyimposeauniforminterchangefee.Thejointdeterminationofinterchangefeeshasgeneratedcontroversyandbeenthefocusoftheeconomicanalysisofthisindustry.

Thereareseveralextensions.Byintroducingconsumersheterogeneity,Dessein(2003)showsthatthepro tneutrality(withrespecttoaccesscharge)undertwo-parttari fails.Thewelfare-optimalaccesschargeisabovethemarginalcost.Inthecontextofasymmetricnetworks,CarterandWright(2003) ndthatitise cienttohavethelargenetworktochoosethereciprocalaccesscharge.Jeon,La ontandTirole(2004)allowreceiverstoderivebene tsfromreceivingcallsanda ectthevolumeofcommunicationsbyhangingup.Whenreceptionchargesaremarket-determined,eachnetworkoperator ndsitoptimaltosetthepricesforcallingandreceptionatitso -netcosts.Thesymmetricequilibriumise cientforanappropriatechoiceofterminationcharge.

10Thereisanothersysteminwhichtheownerofthesystemistheonlyissueroftheassociatedcreditcards;thetwoexamplesareAmericanExpressandDiscover.9

InterchangeFeesandNeutralityinPaymentSystems.Theformalanaly-sisofinterchangefeeswasinitiatedbyBaxter(1983),followinganantitrustlawsuitagainsttheVisaassociation.11Hismaincontributionisaneutralityresultstat-ingthattheleveloftheinterchangefeeisirrelevanttoconsumerprices,providedthatboththeissuersandtheacquirersareperfectlycompetitive.Inamoregeneralanalysis,GansandKing(2003)showthatneutralityholdsregardlessofthelevelofcompetitionbetweenmerchantsandbetween nancialinstitutionsintheabsenceofthesocalledno-surchargerule(NSR).12TheiranalysisimpliesthattheinterchangefeeisgenerallyneutraliftheNSRislifted.Indeedinthiscase,cardusersfacethewholecostofusingacardandhencethereisnoexternality.

AssumingtheNSRtohold,RochetandTirole(2002)letanissuing nancialinstitutionhavemarketpower,theacquiringinstitutionsbeperfectlycompetitiveandmodelmerchantcompetitionalaHotelling.They ndthatthereexistsequi-libriawheremerchantsprefertonotacceptcardpayments—“merchantresistance”—andtheiracceptancepoliciesexhibitstrategiccomplementarity.TheinteractionoftheexternalitygeneratedbytheimpositionoftheNSRwithmerchants’behaviorimpliesthatthepro t-maximizinginterchangefeemaybehigherorlowerthanthewelfare-maximizingone;whenitishigher(lower),theremaybetoomany(few)cardtransactions,respectively.Theythengoontoanalyzethedeterminantsof“merchantresistance”.Inparticular,competitionbetweencardassociationstendstoincrease“merchantresistance”,becauseassociationsnowcompeteformerchantacceptancethroughthemerchantcharge(drivenbytheinterchangefee).However,liftingtheNSRmayincreaseordecreasewelfare.

ByrelaxingtheperfectcompetitionassumptioninBaxter(1983)andassumingimplicitlytheNSRtohold,Schmalensee(2002)attemptstocapturethedouble-sidedexternalityinthesemarketsandarguesthataninterchangefeeisnecessarytobalancecardholderandmerchantdemands.RochetandTirole(2002)andRoger(2002)highlightthecriticalroleoftheno-surchargerule.Indeed,thisruleturnsregular,one-sidedmarketswhereuserscompletelyinternalizetheirconsumptionoftheserviceintotrickiertwo-sidedmarkets,wheredemandfortheserviceononesidegeneratesanexternalityontheotherside.Thistwo-sidedness,whenappliedtothe

NationalBancardCorporation(NaBanco),aspecializedacquirer,suedVisainJune1979,claimingthatVisa’sinterchangefeearrangementsconstitutedaviolationoftheShermanAct.VisawonthebattlebothintheDistrictCourtandintheAppellateCourt.Thecaseendedin1986whentheSupremeCourtdeclinedtoreviewtheAppellateCourt’sdecision.SincethenVisacontinuedtousetheinterchangefeesystem.

12Theno-surchargerule(NSR)isacontractualagreementbetweenmerchantsandtheir(acquir-ing)bankstonotchargedi erentpricestocardusersandothercustomers.11

creditcardindustry,isbestcapturedbyWright(2004).13HeallowsforimperfectcompetitionbetweenacquirerswhilestilllettingtheNSRhold.Thuspositiveexter-nalitiesariseoneachsideofthemarket.Theanalysisfocusesontheoptimalityoftheinterchangefee,regardlessoftheexactformofcompetitionbetweenmerchantsorbetween nancialinstitutions.Duetothetwo-sidedstructureofthismarket,optimalitydependsalsoonthepass-throughrateofcosts,i.e.,theextenttowhichacquirersandissuerscanpassoncostincreases(orretaincostreductions)totheircustomers.Theauthorprovidesanecessaryandsu cientconditionforthewelfare-maximizinginterchangefeetoexceedthepro t-maximizingone.Forinstance,whenmerchantsengageinHotellingcompetitionandissuersandacquirerspassthroughcostsatthesamerate,theconditionissimplythatatthepro t-maximizinginter-changefeetheaveragebene ttoallthosemerchantswhoacceptcardsexceedsthefeetheypay.Raisingtheinterchangefeeincreasesthemerchantfee,whichinducessomemerchantstorejectthecard(increases“merchantresistance”),thusreducingthevolumeoftransaction.ThisdoesnothappenintheRochetandTirole(2002)model,sincetheacquirers’pass-throughrateisuniformlyhigherthanthatoftheissuers,duetotheassumptionofperfectcompetitionamongacquiringinstitutions.Wright(2004)furthermoreestablishesthatundertheaboveconditions,theaverageretailpricesformerchantsacceptingcardswillbehigher.Thisanalysispointsoutthesourceofpotentialdeviationsbetweentheprivatelyandsociallyoptimalleveloftheinterchangefee.

NonexclusiveMembershipinCompetingJointVentures.Aninterest-ingfeatureoftheVisaandMasterCardassociationsisthatamemberofMasterCardcanalsobeamemberofVisa.Thisfeature,knownas“memberduality”,historicallygeneratedsomecompetitionpolicyconcerns.Inarecentpaper,Hausman,Leonard,andTirole(2003)analyzethecompetitiveandgovernancee ectsofthis“duality.”Inparticular,theyinvestigatetheassociations’incentivesforinvestmentsforinnova-tion(cost-reduction)and ndthatinthepresenceofduality,iftheassociationsarenot-for-pro tandcollectusage-basedfeesfrommemberstocovercosts,thentheychooseeconomicallye cientlevelofinvestmentininnovation.Acombinationofnot-for-pro tandusage-basedfeeshelpeliminatethepotentialproblemofexercisingupstreammarketpowerbytheassociations.Ontheotherhand,iftheassociationsarepro t-maximizersorthereisnoduality,theassociationsinvestlessthantheef- cientlevel.Thisanalysismayshedsomelightsonregulatorypoliciestowardjointventuresandstrategicalliancesarisingincreditcards,ATMnetworks,andB2B

RochetandTirole(2004)haveundertakenageneralanalysisoftwo-sidedmarkets,whichwewilldiscussinthenextsection.13

本文来源：https://www.bwwdw.com/article/xpr4.html